EA-SDP link in PMA surveys

(Updated: 2020-06-15)

This presents results of “EA-SDP link” in PMA surveys.

See list of questions/topics on the left panel.
All surveys conducted in select 12 countries have been analyzed (n = 51 surveys). All surveys are publicly accessible and downloaded from PMAdata.org

NOTE: WHY LINKING? 

SDP surveys in PMA are designed to cover SDPs that are either geographically or administratively linked to sampled EAs for the household/female surveys. Thus, when SDP characteristics (e.g., readiness to provide FP service) are linked to the index EAs, we can assess 'population-level accessibility to quality services'.   

It is possible to have other cluster-level aggregate service quality variables using information in female surveys (e.g., cluster mean of MII among users). However, such indicators are reported only among current users. Also, if individual factors determine utilization (e.g., individual demographic and socioeconomic characteristics), rather than cluster-level factors, aggregation of information from only users may be inappropriate to understand associations between service quality and utilization (and later causality using panel data).

1. Overview of the SDP surveys

1.1. Data note

country	refresh_round	note
Burina Faso	5	30 EAs added in R3.
Cote d’Ivoire	none	-
DRC	NONE?	2 EAs dropped in Kinshasa in R2
Ethiopia	5	21 EAs added in R3 (in Oromia).
India, Rajasthan	NONE?	R2: expanded private SDP sample (up-to 3 SDPs in each contiguous urban EAs and from non-selected rural segments or contiguous rural EAs)
Kenya	3 and 5	31 EAs added in R5 (in Nairobi and two new counties: West Pokot and Kakamega)
Niger	NONE?	-
Nigeria, Kano	no refresh	-
Nigeria, Lagos	no refresh	-
Uganda	5	-

1.2. Number of EAs and SDPs in each survey

The number of EAs and SDPs should be roughly stable, unless there was a change (see above). Note:

Burkina Faso Round 6: the number of SDPs decreased. More details in the next section.

1.3. Distribution of SDPs by type in each survey

The public/private composition also should be roughly stable.

Burkina Faso Round 6: It looks like the decrease is primarily public facilities, which should really remain stable.

2. EA-SDP linking methods

EA-SDP link was done using Stata (see the do file in GitHub). Then a summary dataset describing link results was constructed and used for this report.

Briefly, see the following steps:

A. Collapse HHQFQ dataset to EA-level, including survey round, admin level 1, and urban/rural classification for each EA.

B. Reshape SDP dataset to LONG file, using “EAserved_” variables. Now the dataset has unique EA-SDP pairs. An EA should be paired with multiple SDPs, per survey design. A SDP can be paired with multiple EAs.

Technically, all EAs in a survey should appear here, but in some surveys a few EAs are not listed. This is a main (if not the only) reason for EAs not having linked SDPs. This can be caused by various reasons:

The EA truly does not have any SDPs that are “linkable” (unlikely),
It was not possible to survey any of the linked SDPs (due to security reasons, for example), or
Data entry/management errors in “EAserved_” variables in the SDP surveys.

C. Collapse/Create EA-level SDP dataset which has service environment characteristics based on linked SDPs. For example, for each EA, the number of any SDPs offering injectables, the number of primary-level SDPs offering injectables, the number of public SDPs offering injectables, etc.

D. Merge with women datasets (though not necessary for this assessment). If there are missing EAs in above step 2 (and thus 3 too), women in those EAs will be categorized to not have ‘population-level accessibility to quality services’. If the missing EAs are random and minimal, it’s not a big problem.

3. Results

3.1. Number of linked SDPs per EA, by type and level of SDPs

Theoretically expected average number of SDPs linked to an EA:

Any/all SDPs: between 3-6 (3 public + up to three private), but probably slightly above 3. This includes any kind of SDPs, including pharmacies.
Primary or secondary SDPs: between 2-5, but probably slightly above 2. This excludes pharmacies.
Public SDPs: 3, possibly 2 depending on health systems (e.g., Ethiopia). This excludes pharmacies, based on design, since there are no public pharmacies.
Primary or secondary public SDPs: 2, possibly 1 depending on health systems (e.g., Ethiopia). This excludes pharmacies, based on design, since there are no public pharmacies.

More importantly, the average number should remain stable across all rounds, within a geography. Two potential problems (neither related with EA refresh):

Burkina Faso Round 6
Niger, Niamey Round 5
Nigeria, Kano Round 3

3.2. Percent of EAs with no linked SDPs

Theoretically, this should be 0%. But, see:

Burkina Faso Round 6
Niger, Niamey Round 5
Nigeria, Kano Round 3

Other than these two survey, EA-SDP linking looks ‘okay’ (i.e., <5 %) with any SDPs (dark orange bars) and any public SDPs (dark blue bars). For women in these SDP-missing-EAs, any ‘population-level accessibility to quality services’ will be automatically none/missing.

However, percent of EAs without linked primary or secondary SDPs is high. Median (among 51 surveys): 4.5 % for any primary or secondary SDPs (light orange bars) and 11.5 % for public primary or secondary SDPs (light blue bars).

3.3. Percent of EAs with no linked SDPs by residential area

Overall, urban EAs tend to have more problems in linking.

See GitHub for data, code (for both Stata and R), and more information.
For typos, errors, and questions, contact me at yj.choi@isquared.global.

Making Data Delicious, One Byte at a Time